CanCLID: Cantonese Computational Linguistics Infrastructure

CanCLID

The Cantonese Computational Linguistics Infrastructure Development Workgroup (CanCLID) is a collaborative effort to build open resources and tooling for Cantonese language technology.

Contributions

Corpus development

• Collected and curated a high-quality corpus of online text in Guangzhou Cantonese.
• Organized text data for later language-resource and NLP use.

Classification and NLP

• Trained a Cantonese language model based on Guangzhou and Hong Kong text corpora.
• Built a Cantonese/Mandarin classifier from language-model-based classification work.

Mozilla Common Voice localization

• Led Cantonese localization work, including UI translation.
• Worked on corpus collection/refinement, audio recording, and recording verification.

Input method resources

• Optimized the Cantonese IME algorithm to enhance functionality and prediction accuracy.
• Worked on data and usability problems around Cantonese text input.

Methods

• Python and pandas for data processing.
• Corpus cleaning and filtering.
• Git/GitHub collaboration for open language-resource work.

Project Status: Core contributor (2020 - Present) Organization: Cantonese Computational Linguistics Infrastructure Development Workgroup
Repository: github.com/CanCLID